Re: Autohinter project for GSoC

[Werner LEMBERG]

>> As far as I can see it is hard to split the project for adding a
>> capability database to the auto-hinter …
> 
> Just speculating here… might it be possible to reduce the workload
> by only focussing on a few characters at first?  From what I can see
> the single worst Latin-script diacritics with regards to hinting
> seem to be over- and underdots (e.g. ⟨ijḷṃṛ⟩), diaereses, rings,
> tildes and breves (e.g. ⟨ăĕĭŏŭ⟩, cf. caron ⟨ǎěǐǒǔ⟩), so addressing
> just those might provide a disproportionally high benefit.

I think that the development of a generic database format and its
integration into the auto-hinter is the main problem.  The most
important question to answer is how to tell the auto-hinter what to
do, in a rather generic way.

Let's take letter 'i': What exactly is the auto-hinter supposed to do?
There must be some higher-level command(s) in the database like 'the
distance between the upper and lower outline must not become smaller
than (say) 1px'.  There are already 'actions' in `aflatin.c`, for
example, 'ANCHOR', 'ADJUST', 'STEM', etc.; these commands probably
should influence the actions.

Adding more characters is then a probably tedious (due to testing) but
rather trivial issue; this means that adding more characters to the
database is a bonus, to be done *after* implementing a database.


    Werner